A face detection method and system based on multi-level feature deep learning

Abstract

The invention discloses a face detection method and system based on multi-level feature deep learning. The method includes obtaining training set and verification set images and label information of the images; inputting the images and labels into the deep convolutional neural network, training and tuning the network to obtain a face detection model; inputting the test images into the trained The model performs face detection. The neural network includes a first network module for extracting first-level features, a second network module for extracting second-level features, and two branches, which are used to complete face classification tasks and face position coordinates respectively The third network module for the regression task. The present invention detects on feature maps of different levels, solves the problems of slow efficiency, cumbersome steps and poor generalization ability in the existing method of adopting the staged method, and achieves real-time detection while ensuring high detection accuracy Effect, and has good robustness to the illumination and occlusion problems of face images.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a face detection method and system based on deep learning of multi-level features. Background technique [0002] Face detection refers to searching any input image to determine whether it contains a human face, and if it contains a human face, it returns information such as the position and size of all faces. It is the premise and foundation of face information processing, and the quality of detection performance directly affects the result of face information processing. Face detection is very simple for people, but the image stored in the machine is just a binary string composed of 0 and 1, an array of numbers, so for the machine, face detection is very complicated. Task. The main goal of machine detection of faces is to ensure a high accuracy rate under the premise of low calculation. The research on face detection stems from the problem of face recognition. The earliest fac...

AI Technical Summary

Problems solved by technology

[0005] Purpose of the invention: In view of the deficiencies in the prior art, the purpose of the present invention is to provide a face detection method and system based on multi-level feature deep learning, which solves the existing human face detection problem by performing detection on different levels of convolution output feature maps. The face detection algorithm has the disadvantages of cumbersome training steps, slow efficiency and poor network generalization ability

Images